Loading...

Media is loading
 

Degree

Bachelor of Science (Computer Science)

Department

Department of Computer Science

School

School of Mathematics and Computer Science (SMCS)

Advisor

Muhammad Zain Uddin, Lecturer, Computer Science-SMCS

Co-Advisor

Muhammad Adil Saleem

Keywords

AI-Powered Business Intelligence, Intelligent Data Analytics, Semantic Data Querying, Local-First Analytics

Abstract

This project presents an AI-powered Business Intelligence and Analytics platform designed to enable intelligent querying, automated insight generation, and secure semantic analysis of enterprise data. The system supports data ingestion through PostgreSQL and MySQL database connectors. To minimize production database load and improve performance, selected metadata and sampled records are fetched and stored locally using DuckDB for offline analytical processing. The platform employs semantic modelling techniques to construct a semantic layer containing relationships, schema mappings, KPI definitions, and contextual business metadata. Retrieval-Augmented Generation (RAG) combined with vector databases is used to provide contextual understanding and improve Natural Language Processing (NLP)-based query interpretation. Users can interact with the system using natural language queries, which are converted into optimized SQL or DAX queries through NLP-to-SQL/DAX generation models. The solution provides advanced analytics capabilities including executive KPI reporting, trend and seasonality analysis, anomaly detection. Large Language Models generate explainable summaries and insights grounded in retrieved semantic context. The system can also automatically generate charts and lightweight dashboards based on user intent. A local-first deployment architecture is adopted to ensure enterprise data privacy and security by keeping all sensitive organizational data within the client environment. The proposed system aims to simplify business analytics workflows, reduce dependency on manual dashboard creation, and enhance decision-making through AI-driven insights and semantic understanding of data.

Tools and Technologies Used

Python, FastAPI, Node.js, React, PostgreSQL, MySQL, DuckDB, SQL, DAX, NLP-to-SQL, NLP-to-DAX, Retrieval-Augmented Generation (RAG), Vector Database, Semantic Modelling, Metadata Extraction, Machine Learning Models, Large Language Models (LLMs), Pandas, NumPy, REST APIs, Embedding Models, Local Storage Processing, Data Cleaning & Preprocessing, Automated Query Generation, KPI Analytics, Trend & Anomaly Detection, Dashboard Visualization, Git, Cloud APIs

Methodology

The project follows an AI-driven Business Intelligence and Analytics methodology focused on secure local data processing and semantic understanding of enterprise data. The system first ingests data through direct PostgreSQL or MySQL database connectors. For SQL databases, metadata and selected top-N rows are fetched initially to reduce latency and avoid heavy production database load. Required tables and columns are identified using metadata-driven recommendation models, after which the selected data is stored locally in DuckDB for offline querying and processing.

The ingested data then undergoes preprocessing and cleaning before semantic modelling techniques are applied to create a semantic layer containing relationships, schemas, KPI definitions, column mappings, and contextual business meanings. 

The semantic layer is integrated with a Retrieval-Augmented Generation (RAG) pipeline and vector database to enable natural language querying. User queries are embedded and matched against contextual metadata and business rules to optimize prompt generation and improve NLP-to-SQL/DAX conversion accuracy. Query generation models dynamically generate optimized SQL or DAX queries to retrieve relevant information.

The analytics engine performs KPI analysis, trend detection, anomaly detection, using LLM. The Large Language Models generate explainable insights, summaries, and recommendations grounded in retrieved business context. The system can additionally generate charts, KPIs, and lightweight dashboards automatically based on the query intent.

The entire application is designed as a downloadable local deployment solution to ensure user data privacy and prevent sensitive enterprise data from leaving the client environment.

Document Type

Restricted Access

Submission Type

BSCS Final Year Project

Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Download

Share

COinS